Implement Benchmarks

[TobyBoyne]

We have a long list of benchmarks we'd like to implement. They are all listed below:

(Semi-) Synthetic

• SVM opt [1]
• Possibly Antibody Design and RNA from [1]
• Welded beam [2]
• Cellular Network [2]
• Crossed Barrel [3]

Neural Architecture Search

• We want to implement queryable NAS problems [4] - that is, we don't want to have to train a NN at every time step as this makes the experiments much more costly, so we want a precomputed database/surrogate where we can just evaluate at (almost) arbitrary points. Need to investigate how these are implemented.

[1] Dreczkowski et al 2023 | https://github.com/huawei-noah/HEBO/tree/master/MCBO

[2] Daulton et al 2022 | https://github.com/facebookresearch/bo_pr

[3] Zhu et al 2024 | https://github.com/MolChemML/ExpDesign

[4] White et al 2023

[TobyBoyne]

Also: XGBoostMNIST